Backlight control and display mapping for high dynamic range images

Abstract

Systems and methods are disclosed for dynamically adjusting the backlight of a display during video playback. Given an input video stream and associated minimum, average, or maximum luminance values of the video frames in the video stream, values of a function of the frame min, mid, or max luminance values are filtered using a temporal filter to generate a filtered output value for each frame. The instantaneous dynamic range of a target display is determined based on the filtered output value and the minimum and maximum brightness values of the display. A backlight control level is computed based on the instantaneous dynamic range, and the input signal is tone mapped by a display management process to be displayed on the target display at the selected backlight level. The design of a temporal filter based on an exponential moving average filter and scene-change detection is presented.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority to U.S. Provisional Patent Application No. 62/160,353, filed on May 12, 2015, U.S. Provisional Patent Application No. 62/193,678, filed on Jul. 17, 2015 and European Patent Application No. 15171033.2, filed on Jun. 8, 2015, each of which is incorporated herein by reference in its entirety.

TECHNOLOGY

The present invention relates generally to images. More particularly, an embodiment of the present invention relates to backlight control and display mapping for images and video sequences with high dynamic range.

BACKGROUND

As used herein, the term ‘dynamic range’ (DR) may relate to a capability of the human visual system (HVS) to perceive a range of intensity (e.g., luminance, luma) in an image, e.g., from darkest darks (blacks) to brightest whites (highlights). In this sense, DR relates to a ‘scene-referred’ intensity. DR may also relate to the ability of a display device to adequately or approximately render an intensity range of a particular breadth. In this sense, DR relates to a ‘display-referred’ intensity. Unless a particular sense is explicitly specified to have particular significance at any point in the description herein, it should be inferred that the term may be used in either sense, e.g. interchangeably.

As used herein, the term high dynamic range (HDR) relates to a DR breadth that spans the some 14-15 orders of magnitude of the human visual system (HVS). In practice, the DR over which a human may simultaneously perceive an extensive breadth in intensity range may be somewhat truncated, in relation to HDR. As used herein, the terms enhanced dynamic range (EDR) or visual dynamic range (VDR) may individually or interchangeably relate to the DR that is perceivable within a scene or image by a human visual system (HVS) that includes eye movements, allowing for some light adaptation changes across the scene or image. As used herein, EDR may relate to a DR that spans 5 to 6 orders of magnitude. Thus while perhaps somewhat narrower in relation to true scene referred HDR, EDR nonetheless represents a wide DR breadth and may also be referred to as HDR.

In practice, images comprise one or more color components (e.g., luma Y and chroma Cb and Cr) wherein each color component is represented by a precision of n-bits per pixel (e.g., n=8). While SDR images can typically be encoded with 8-10 bits per color component, EDR and HDR images typically require more than 8 bits (e.g., 10-12 bits, or more). EDR and HDR images may also be stored and distributed using high-precision (e.g., 16-bit) floating-point formats, such as the OpenEXR file format developed by Industrial Light and Magic.

A reference electro-optical transfer function (EOTF) for a given display characterizes the relationship between color values (e.g., luminance) of an input video signal to output screen color values (e.g., screen luminance) produced by the display. For example, ITU Rec. ITU-R BT. 1886, “Reference electro-optical transfer function for flat panel displays used in HDTV studio production,” (March 2011), which is included herein by reference in its entity, defines the reference EOTF for flat panel displays based on measured characteristics of the Cathode Ray Tube (CRT). Given a video stream, any ancillary information is typically embedded in the bit stream as metadata. As used herein, the term “metadata” relates to any auxiliary information that is transmitted as part of the coded bitstream and assists a decoder to render a decoded image. Such metadata may include, but are not limited to, color space or gamut information, reference display parameters, and auxiliary signal parameters, as those described herein.

Most consumer HDTVs range from 300 to 500 nits with new models reaching 1000 nits (cd/m²). As the availability of HDR content grows due to advances in both capture equipment (e.g., cameras) and displays (e.g., the PRM-4200 professional reference monitor from Dolby Laboratories), HDR content may be color graded and displayed on displays that support higher dynamic ranges (e.g., from 1,000 nits to 5,000 nits or more). Such displays may be defined using alternative EOTFs that support high luminance capability (e.g., 0 to 10,000 nits). An example of such an EOTF is defined in SMPTE ST 2084:2014 “High Dynamic Range EOTF of Mastering Reference Displays,” which is incorporated herein by reference in its entirety. In general, without limitation, the methods of the present disclosure relate to any dynamic range higher than SDR. As appreciated by the inventors here, improved techniques for the display of high-dynamic range images are desired.

The approaches described in this section are approaches that could be pursued, but not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated, it should not be assumed that any of the approaches described in this section qualify as prior art merely by virtue of their inclusion in this section. Similarly, issues identified with respect to one or more approaches should not assume to have been recognized in any prior art on the basis of this section, unless otherwise indicated.

BRIEF DESCRIPTION OF THE DRAWINGS

An embodiment of the present invention is illustrated by way of example, and not in way by limitation, in the figures of the accompanying drawings and in which like reference numerals refer to similar elements and in which:

FIG. 1 depicts an example process for backlight control and display management according to an embodiment of this invention;

FIG. 2 depicts an example relationship between a backlight setting and the instantaneous dynamic range for a display with dynamic range control;

FIG. 3 depicts an example process for adaptive luminance range mapping according to an embodiment; and

FIG. 4A and FIG. 4B depict examples of applying moving average filtering to luminance values of an input sequence according to embodiments of this invention.

DESCRIPTION OF EXAMPLE EMBODIMENTS

Techniques for backlight control and display management or mapping of high dynamic range (HDR) images are described herein. In the following description, for the purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, that the present invention may be practiced without these specific details. In other instances, well-known structures and devices are not described in exhaustive detail, in order to avoid unnecessarily occluding, obscuring, or obfuscating the present invention.

Overview

Example embodiments described herein relate to backlight control and display management of HDR images. In an embodiment, given an input video stream and one or more of associated minimum (min), average (mid), or maximum (max) luminance values of the video frames in the video stream, values of a function of the frame min, mid, or max luminance values are filtered using a temporal filter to generate a filtered output value for each frame. The instantaneous dynamic range of a target display (e.g., as defined by a minimum and maximum luminance values of the display at the desired brightness) is determined based on the filtered output value and the minimum and maximum brightness values of the display. A backlight control level is computed based on the instantaneous dynamic range and the input signal is tone mapped by a display management process to generate a tone-mapped output to be displayed within the selected instantaneous dynamic range. The design of a temporal filter based on an exponential moving average filter and scene-change detection is presented.

In an embodiment, the filtered min, mid, or max luminance values for the input video stream are received through signal metadata.

In an embodiment, the filtered min, mid, or max luminance values for the input video stream are computed by an image analysis block.

In an embodiment, the filtered output values for each frame of the input video stream and/or the computed instantaneous dynamic range for a target display may be stored as metadata to be transmitted together with the tone mapped output.

Example Display Control and Display Management

FIG. 1 depicts an example process (100) for display control and display management according to an embodiment. Input signal (102) is to be displayed on display (120). Input signal may represent a single image frame, a collection of images, or a video signal. Image signal (102) represents a desired image on some source display typically defined by a signal EOTF, such as ITU-R BT. 1886 or SMPTE ST 2084, which describes the relationship between color values (e.g., luminance) of the input video signal to output screen color values (e.g., screen luminance) produced by the target display (120). The display may be a movie projector, a television set, a monitor, and the like, or may be part of another device, such as a tablet or a smart phone.

Process (100) may be part of the functionality of a receiver or media player connected to a display (e.g., a cinema projector, a television set, a set-top box, a tablet, a smart-phone, a gaming console, and the like), where content is consumed, or it may be part of a content-creation system, where, for example, input (102) is mapped from one color grade and dynamic range to a target dynamic range suitable for a target family of displays (e.g., televisions with standard or high dynamic range, movie theater projectors, and the like).

In some embodiments, input signal (102) may also include metadata (104). These can be signal metadata, characterizing properties of the signal itself, or source metadata, characterizing properties of the environment used to color grade and process the input signal (e.g., source display properties, ambient light, coding metadata, and the like).

In some embodiments (e.g., during content creation), process (100) may also generate metadata which are embedded into the generated tone-mapped output signal. A target display (120) may have a different EOTF than the source display. A receiver needs to account for the EOTF differences between the source and target displays to accurate display the input image. Display management (115) is the process that maps the input image into the target display (120) by taking into account the two EOTFs as well as the fact that the source and target displays may have different capabilities (e.g., in terms of dynamic range.)

As used herein, the terms “display management” or “display mapping” denote the processing (e.g., tone and gamut mapping) required to map an input video signal of a first dynamic range (e.g., 1000 nits) to a display of a second dynamic range (e.g., 500 nits). Examples of display management processes can be found in U.S. Provisional Patent Application Ser. No. 62/105,139 (to be referred to as the '139 application), filed on Jan. 19, 2015, titled “Display management for high dynamic range images,” also filed as PCT Application Ser. No. PCT/US2016/013352, on Jan. 14, 2016, each of which is incorporated herein by reference in its entirety.

In some embodiments, the dynamic range of the input (102) may be lower than the dynamic range of the display (120). For example, an input with maximum brightness of 100 nits in a Rec. 709 format may need to be color graded and displayed on a display with maximum brightness of 1,000 nits. In other embodiments, the dynamic range of input (102) may be the same or higher than the dynamic range of the display. For example, input (102) may be color graded at a maximum brightness of 5,000 nits while the target display (120) may have a maximum brightness of 1,500 nits.

In an embodiment, display (120) is controlled by display controller (130). Display controller (130) provides display-related data (134) to the display mapping process (115) (such as: minimum and maximum brightness of the display, color gamut information, and the like) and control data (132) for the display, such as control signals to modulate the backlight or other parameters of the display for either global or local dimming. An example of a display controller for dual modulation display systems is described in U.S. Pat. No. 8,493,313, “Temporal filtering of video signals,” by G. Damberg and H. Seetzen, which is incorporated herein by reference in its entirety. Another example is described in PCT Application Ser. No. PCT/US2014/012568 (WO 2014/116715A1), filed on Jan. 22, 2014, “Global display management based light modulation,” by T. Kunkel, which is incorporated herein by reference in its entirety.

Displays using global or local backlight modulation techniques adjust the backlight based on information from input frames of the image content and/or information received by local ambient light sensors. For example, for relatively dark images, the display controller (130) may dim the backlight of the display to enhance the blacks. Similarly, for relatively bright images, the display controller may increase the backlight of the display to enhance the highlights of the image. For example, FIG. 2 depicts a display with a full dynamic range of Max_BL/Min_BL, where Max_BL denotes the maximum possible luminance of the display when the backlight is full on (e.g., 1,000 nits), and Min_BL denotes the minimum possible luminance of the display when the backlight is full off (e.g., 0.001 nits). Note that, for illustration purposes, the Y axis of FIG. 2 depicts log luminance values. As used herein, the term “instantaneous dynamic range” denotes a simultaneous (or static) dynamic range, that is the range of black (e.g., Min_T (K)) to white (e.g., Max_T(K)) that can be shown simultaneously on a display for a given backlight level (K) at some instance of time. Let Max_T(K)/Min_T (K) (e.g., 1,000:1) denote the instantaneous (also referred to as static) dynamic range of the display (120) for a specific level K (205) of backlight between full off and full on, then display management (115) maps the dynamic range of the input (102) into the range (Min_T(K), Max_T(K)) of the display. The instantaneous dynamic range ratio Max_T(K)/Min_T (K) may be fixed or approximately the same for all values of K. For example, for an instantaneous dynamic range of 1,000:1, w=log(1,000)/2=1.5.

In an embodiment, the display mapping (115) and display control (130) processes are enhanced by suitable image analysis (105) and image processing (110) operations as will be described herein.

Image Analysis

In an embodiment, unless specified already by the source metadata (104), for each input frame in signal (102) the image analysis (105) block may compute its minimum (min), maximum (max), and median (mid) (or average gray) luminance value. These values may be computed for the whole frame or part of a frame. In some embodiments, min, mid, and max luminance values may represent approximate values of the true values. For example, computed min and max values may represent 90% of the true min and max values in the input signal so as to be more robust to single pixel outliers.

In some embodiment, min, mid, and max luminance signal values may also be computed or received as metadata for a whole scene. As used herein, the terms ‘scene’ or ‘shot’ denote a series of sequentially-in-time captured sequence frames that may share the same overall color or brightness characteristics. Authored content, such as films and pre-recorded video can be edited in such a way that image statistics may be computed over a cohesive set of frames, such as a scene or a “cut,” which may prevent temporal artifacts; however, in computer games and live broadcast, there might not be enough information to have pre-determined scene cuts, so better adaptation techniques are required.

Scene cuts may be determined automatically or they may be denoted in the bitstream using metadata. Automatic scene change detection is a challenging and well-studied problem. Embodiments of this invention can easily tolerate missed scene cuts or false detected scene cuts, hence the exact method of scene-cut detection is not particularly important; nevertheless, without limitation, a variety of scene cut detection mechanisms are suggested herein.

For example, let Mid_S1 and Mid_S2 denote respectively the mid luminance values for two consecutive scenes S1 and S2, then, in an embodiment a scene cut may determined if:
Mid_S1−Mid_S2≠0,  (1)
or
|Mid_S1−Mid_S2|>T_F,
where T_F is a predefined threshold (e.g., T_F=0.1).
Target Display Luminance Range Mapping

Given min, mid, and max luminance source data (107 or 104), image processing block (110) may compute the display parameters (e.g., Min_T and Max_T, or the level K of backlight) that allow for the best possible environment for displaying the input video. Due to brightness fluctuations even within frames in the same scene, treating each frame independently may lead to flickering and other unwanted visual artifacts. In an embodiment, a temporal filter is applied to a sequence of sequential frames in the scene to determine the best luminance mapping (e.g., Min_T and Max_T). In a preferred embodiment, luminance mapping employs a temporal filter based on an exponential moving average (EMA) filter; however, other FIR or IIR temporal filters could be applied as well. In some embodiments, temporal filtering and other aspects of luminance range mapping (110) may applied at the source display, and the filter output data may be passed to the target display as metadata. This allows for fewer computations at the target display and additional creative control by the content provider. For example, the content creator (e.g., a director or a color grader) may decide to override the results of the filter output (110) to manually adjust how the image is displayed.

Let L_F(t) denote a function of min, mid, and max luminance values in a frame at time t in a scene. In an embodiment L_F(t) may be simply the mid luminance value of a frame at time t in a scene (e.g., L_F(t)=Mid_F(t)). In other embodiments, L_F(t) may represent the min or max values, or a weighted combination of the min, mid, and max values. Then, in an embodiment, EMA filtering in a scene may be expressed as:
S₀=L_F(0), for t=0,
S_t=α*L_F(t)+β*S_t-1, for t>0,  (2)
where α (alpha) and β (beta) denote weight factors.

In an embodiment,
β=1−α.
In some embodiments, the weights may be fixed (e.g., α=0.25, β=0.75). In some embodiments the weights may be selected from a fixed list of possible weights. For example, for L_F(t)=Mid_F(t) the alpha value may be fixed (e.g. α=0.25), but for L_F(t)=Max_F(t) and L_F(t)=Min_F(t) the value of alpha may switch between two or more values, say α₁=0.175 and α₂=0.475. This will be referred to as asymmetric alpha. For example, in an embodiment that uses two asymmetric alpha values, if S_t>S_t-1, then for the next data point α=α₂, otherwise α=α₁. This allows tone-mapping operations to adapt quicker to new increased highlights or lower darks in the input image sequences.

FIG. 4A depicts examples of EMA filtering for the min (405-a), mid (410-a), and max (415-a) luminance values for about 200 frames in a video sequence. Filtering of the mid value uses a fixed alpha, while filtering of the min and max values uses asymmetric alpha values, as described earlier.

In some embodiments the weights may be a function of the delivery frame rate. Such an implementation is especially important for video streaming applications where the frame rate may change dynamically based on either computational resources or available bandwidth. For example, if α_m denotes a weight factor optimized for a delivery at M frames per second (e.g., M=24), and R denotes the actual delivery rate (e.g., R=30), then using a linear conversion:

$\alpha ={\alpha }_M*\frac{M}{R},$
which allows alpha values to decrease when the actual frame rate increases.

In some embodiments β may be defined to be a function of time. For example, in an embodiment:

$\begin{array}{cc}\beta =0,\mathrm{for}t=0\beta =\mathrm{clip}3(0,\frac{1-\alpha }{t},\left(1-\alpha \right),\mathrm{for}0<t\le m,\beta =\mathrm{clip}3\left(0,\frac{1-\alpha }{m},\left(1-\alpha \right)\right),\mathrm{for}t>m,& \left(3\right)\end{array}$
where m>0 is a predetermined time instant and clip3(a,f(x),c) denotes that the output of f(x) is always clipped to be within the values of a and c, where a and c are included.

In some embodiments, the alpha value of the EMA filter may be reset or adjusted when a new scene cut or scene change is detected. For example, in an embodiment:
α=min(1,SceneCut*|S_t-1−L_F(t)|*α_scene+a_base),  (4a)
where SceneCut is in the range (0, 1) and denotes the confidence (or probability) in detecting a scene cut. For example, SceneCut=1 may specify there is a new scene with full confidence. Parameters α_scene and α_base denote predefined filter parameters that control how fast the filter adapts. In an embodiment, without limitation, typical ranges for these variables include a_scene=(2.0, 5.0) and a_base=(0.02, 0.2) (e.g., a_scene=3.0 and α_base=0.05). Hence, when a new scene is detected, the value of a may be increased proportionally to the change of the scene-related metadata (e.g., the average scene luminance) to make smoother the transition between the adjustment in mid brightness values. In some embodiments, in equation (4), S_t-1 may also be substituted with L_F(t-1). From equations (2)-(4), when a new scene is detected, a is getting very close to one and the value of is close to zero, hence, the current L_F(t) values are weighted more than past filtered values. In addition, when a new scene cut is detected, t may be reset to 0, and all of the previous S_t values may be cleared from the memory. In other words, optionally, the memory of the temporal filter may be reset to zero every time there is a scene cut.

As an example, FIG. 4B depicts examples of EMA filtering for the min (405-b), mid (410-b), and max (415-b) luminance values for about 200 frames in a video sequence (same as the one depicted in FIG. 4A) when taking into consideration when a scene-cut is detected.

As depicted in FIG. 4B, scenes (e.g., S1 to S4) in the video sequence may also be characterized and labeled in terms of their statistical features. For example, scene S4 may be determined to be statistically equivalent to scene S2 in terms of average min, mid, and max values or of a function of them (e.g., both are labeled as scene “B”). For example, from an implementation point of view, the image analysis unit (315) may keep a scene history log and associated statistics (e.g., averages of min, mid, max, and S_t values). This type of characterization may be used for a far more effective initialization of the EMA filter in equation (2) than always starting with the default (e.g., S₀=L_F(0), for t=0). For example, one may initiate the EMA filter based on statistical data from a previous scene that bears the strongest resemblance to the new scene. For example, for the sequence of scenes in FIG. 4B, at the start of scene S4, one may have S₀=L_S2, for t=0, where L_S2 represents a suitable value computed based on the statistics of scene S2, which is considered to be the closest to scene S4.

In another embodiment, α may be defined as a function of time (α_t). For example,
α_t=α_def+(α_max−a_def)SceneCut,  (4b)
where a_def denotes a default value, α_max denotes a maximum value, and as before, SceneCut denotes a probability of the frame to belong to a new scene. This allows again for faster, but smoother, adaptation of the EMA filter to scene cuts or sudden changes to the luminance values of an input picture. If a frame has low probability to belong to a scene cut, then the default alpha parameter is being used, otherwise, for definite scene cuts, an alpha value closer to the α_max value is being used.

Let Min_S, Mid_S, and Max_S denote the brightness characteristics of a source or reference display, and let Min_T, Mid_T, and Max_T denote the brightness characteristics of the target display (120), then, as described by A. Ballestad et al., in U.S. Pat. No. 8,593,480, titled “Method and apparatus for image data transformation,” which is incorporated herein by reference in its entirety, these values may define the anchor points of a sigmoid-like, tone-mapping function, which together with other tone-mapping operations (e.g., as described in the '139 application) enable the display management process (115) to generate a tone-mapped output (117) to be displayed on the target display (120).

In an embodiment, given the results of the temporal filter (e.g., equation (2)), the preferred instantaneous luminance range for the target display (120) may be computed as
Max_T=clip3(Min_BL,f_max(S_t),Max_BL),  (5)
Min_T=clip3(Min_BL,f_min(S_t),Max_BL),
where f_max(S_t), and f_min(S_t), denote functions to determine the max and min values of the preferred instantaneous dynamic range of the target display based on one or more limit luminance values for the target display (e.g., Min_BL, Max_BL). For example, without limitation, assuming all display luminance values and S_t are in expressed in a linear domain (shown with an overbar) (e.g., Max_BL=10^MaxBL when Max_BL is in log 10 domain), let w denote one half of the instantaneous dynamic range in the linear domain (e.g., see FIG. 2). Then, if

$\begin{array}{cc}{S}_t=\mathrm{clip}3\left(\stackrel{\_}{{\mathrm{Min}}_{\mathrm{BL}}}*\stackrel{\_}{w},S_t,\frac{\stackrel{\_}{{\mathrm{Max}}_{\mathrm{BL}}}}{\stackrel{\_}{w}}\right),\mathrm{then}\stackrel{\_}{{\mathrm{Max}}_T}=\stackrel{\_}{f_{\max }}\left(S_t\right)=S_t*\stackrel{\_}{w},\stackrel{\_}{{\mathrm{Min}}_T}=\stackrel{\_}{f_{\min }}\left(S_t\right)=S_t\text{/}\stackrel{\_}{w}.& \left(6\right)\end{array}$

If S_t values are computed in a gamma or other perceptually-quantized luminance space, then they may have to be linearized first. Alternatively, equations (5)-(6) may also be computed in a logarithmic domain. For example, assuming all luminance values are expressed in logarithmic space, let w in denote one half of the instantaneous dynamic range in the logarithmic domain. Then if
log(S_t)=clip3(Min_BL+w,log(S_t),Max_BL−w),
then
Max_T=f_max(S_t)=log(S_t)+w,
Min_T=f_min(S_t)=log(S_t)−w.  (7)
For example, let a display have Max_BL=1,000 nits, Min_BL=0.001 nits, and an instantaneous dynamic range of 1,000:1. Then Max_BL=3, Min_BL=−3, w=log(1,000)/2=1.5, and w=√{square root over (1000)}=31.62.

Given the Min_T and Max_T values (111) computed by equations (6) or (7), the display controller (130) may then apply a look-up table or other internal-control processes to determine the appropriate level K for controlling the display's backlight. Alternatively, in a content-creation environment, S_t-related values or one or more of the computed Min_T and Max_T values or a function of these values (e.g., Mid_T) may be embedded as metadata in the tone-mapped bitstream to be delivered downstream to content consumers. Hence, a receiver with low computational resources, such as a tablet or a smartphone, may use directly these values to determine the best display setup.

FIG. 3 depicts an example process for adjusting the instantaneous luminance range mapping according to an embodiment. Given input (102) and associated metadata (104), step (310) determines whether luminance-related metadata values (e.g., min, med, and max) for each frame are available. If there are no relevant metadata, then these values may be computed in step (315). In some embodiments, computing luminance-related metadata in (315) may also require a luminance linearization step, where a gamma or other non-linear mapping (as specified by the source's EOTF) is removed to generate data with linear luminance. In some embodiments, the results of image analysis (315) may also be stored and/or passed downstream as signal-related metadata (304). Next, in step (345), it is determined whether the current frame starts a new scene. If there is a scene change, optionally, in step (340), the temporal filter may also be fully reset (e.g., the memory of all past S_t filter output values is cleared and t is set to zero). Next, in step (325), the parameters of the temporal filter are adjusted as needed (e.g., equations (3) and (4)) and the process continues with the temporal filtering (320) (e.g., equation (2)) of values of the L_F(t) function of these metadata values. In step (330), the results (322) of the temporal filtering (e.g., S_t) are used to determine the proper instantaneous luminance range (111) (e.g., Min_T and Max_T) to display the input data (102) on the target display. Appropriate data (111) are passed to the display controller (130) and the display management process (115). Optionally, related metadata (304) (e.g., the filtered L_F(t) values for each frame) may be saved to be stored together with the modified (tone-mapped) by the display management bitstream to be passed to other downstream components.

While example embodiments have been presented for optimizing the display of images (either of standard dynamic range (SDR) or high dynamic range) on high-dynamic range displays, the same techniques may also be applied to improve SDR displays. For example, when viewing a display under high ambient light (e.g., a tablet or smartphone in day light), the techniques may be used to compensate for the low dynamic range caused by the high ambient light and the display's reflectivity parameters.

Preview Frames and Forward and Reverse EMA Filtering

If one can have a preview of the next N frames of the source material, then the image analysis unit (315) may apply this information to improve scene-change detection performance, statistics gathering, and filter initialization. In an embodiment, one may compute the moving average of equation (2) in both time-forward and time-reverse order on the upcoming frames, facilitating detection of when a significant change in scene content takes place. For example, one may compute:
S₀^f=L_F(0), and
S₀^r=L_F(N), for t=0,
S_t^f=α*L_F(t)+(1−α)*S_t-1^f, and
S_t^r=α*L_F(N-t)+(1−α)*S_t-1^r, for t>0.  (8)

That is, in the time-reverse EMA filter (S_t^r), future preview frames are added one-by-one, starting from the most future one (N) and working backwards towards the current frame. If the time-reverse moving average (S_t^r) is getting closer to the time-forward moving average (S_t^f), then one can determine that there is no scene change moment. Likewise, if the two moving averages are within a distance threshold of each other, then one may continue normal in-scene progression; however, when one detects a large difference between the forward and reverse moving average metrics, then one with high confidence can determine there is a scene cut between the current frame and the N-th future frame, typically occurring at the maximum of their distance (e.g. |S_t^f−S_t^r| and the like).

When such a scene change is detected, say at frame k, as described earlier, one can start the EMA filter for the new scene based on the current frame (e.g., S₀^f=L_F(k)). Alternatively, one may choose to initialize the EMA filter based on the already computed EMA values of the preview frames (e.g., S₀^f=S_N-k^r). This approach represents an improvement over starting fresh with only one frame's statistics, and in most cases outperforms the cached history approach described earlier.

Improved Scene-Change Detection

In another embodiment, a scene-cut detection technique is based on statistical differences between the current frame (or a collection of future preview frames) and an existing moving average characteristic. For example, the following statistics may be computed for each input frame:

• a) A 1-D log-luminance histogram (say, h_i^Y) representing a histogram of log-luminance values (e.g., log(Y) of a frame in the YCbCr domain)
• b) A 2-D CIE (u′v′) chromaticity histogram (say, h_j^uv), and
• c) An edge strength histogram (say, h_k^e)

The log-luminance histogram serves both as a measure of scene content and as a resource for tone-mapping in display management. The two-dimensional CIE (u′v′) chromaticity histogram is less common, but serves here to summarize the color content of a scene. The edge strength histogram represents a representation of a frame in terms of its edge content. An example of computing edge-strength histograms is described in Lee, Seong-Whan, Young-Min Kim, and Sung Woo Choi. “Fast scene change detection using direct feature extraction from MPEG compressed videos,” in Multimedia, IEEE Transactions” on 2.4 (2000): 240-254, which is incorporated herein by reference in its entirety. In an embodiment, an edge-strength histogram may be computed as follows:

• a) Take the luminance plane of the input picture (say, Y)
• b) Subsample it, to reduce its resolution, say, by a factor of m in each dimension, e.g., (m=6) to generate Yr. These two steps allow one to reduce the computational complexity of this algorithmic step at the expense of some loss in accuracy.
• c) Apply an edge kernel filter to Yr (say, the Sobel edge filter), to generate an edge map (say, Yr^e)
• d) Compute a histogram of the edge map (say, h_k^e)

In an embodiment, a weighted Euclidian distance (say, Δ_t) between each of these histograms from the current frame (or preview frames) and the moving average of the previous frames is computed to determine whether a scene cut is appropriate at a given point in time. For example, in an embodiment, let
Δt=w₁Δ_t^Y+w₂Δ_t^uv+w₃Δ_t^e,  (9)
where w_i, for i=1, 2, and 3, denotes the weighting factors (e.g., w₁=w₂=0.35 and w₃=0.3), Δ_t^Y denotes a measure of the distance (e.g., L1, L2, and the like) between the luminance histograms of the current frame (e.g., at time t) and a previous frame (e.g., at time t−1) (e.g., using the L2 distance, Δ_t^Y=Σ_i(h(t)_i^Y−h(t−1)_i^Y)²), Δ_t^uv denotes a corresponding distance of the chroma histograms, and Δ_t^e denotes the distance between the corresponding edge histograms. In some embodiments, histogram values in each histogram are filtered by a low-pass filter before computing distance values to improve scene-cut detection.

Given Δ_t from equation (9), a scene cut may be determined if Δ_t is larger than a predetermined threshold. Alternatively, the probability of a scene cut may be determined as a function of Δ_t. For example, the variable SceneCut in equations (4a) and (4b) may be determined as

$\begin{array}{cc}\mathrm{SceneCut}=\frac{{\Delta }_t}{{\Delta }_t+C},& \left(10\right)\end{array}$
where c is a tunable constant.
Enumerated Example Embodiments

Various aspects of the present disclosure may be appreciated from the following enumerated example embodiments (EEEs):

EEE 1. An apparatus configured to:

obtain a stream of values, each of the values being indicative of at least one luminance characteristic of a respective frame of a stream of video data frames, the stream of video data frames comprising a sequence of scenes, each one of the scenes comprising a respective one or more of the frames;

apply a temporal filter to the stream of values to thereby smooth the stream of values; and

output the smoothed stream of values, and further configured to detect a transition from a current to a next one of the scenes, and, in response, to reset the temporal filter.

EEE 2. The apparatus of EEE 1 wherein the temporal filter is configured to store in memory one or more prior values of the stream of values, and wherein resetting the temporal filter comprises clearing the one or more prior values from the memory.

EEE 3. The apparatus of EEE 1 or EEE 2 further configured obtain the stream of video data frames, wherein detecting the transition comprises applying an automatic scene change detection algorithm to the stream of video data frames.

EEE 4. The apparatus of EEE 1 or EEE 2 further configured obtain a stream of metadata associated with the stream of video data frames, wherein detecting the transition comprises obtaining a signalling of the transition from the metadata.

EEE 5. The apparatus of any one of EEE 1 to EEE 4 wherein the at least one luminance characteristic comprises one of, a function of any two of, or a function of all three of: a min luminance value; a mid luminance value; and a max luminance value.

EEE 6. The apparatus of any one of EEE 1 to EEE 5 further configured to use at least the smoothed stream of values to generate a stream of dynamic range values for a target display.

EEE 7. The apparatus of EEE 6 further configured to use at least the dynamic range values to generate a backlight-level control signal for the target display.

Example Computer System Implementation

Embodiments of the present invention may be implemented with a computer system, systems configured in electronic circuitry and components, an integrated circuit (IC) device such as a microcontroller, a field programmable gate array (FPGA), or another configurable or programmable logic device (PLD), a discrete time or digital signal processor (DSP), an application specific IC (ASIC), and/or apparatus that includes one or more of such systems, devices or components. The computer and/or IC may perform, control, or execute instructions relating to backlight control and display mapping processes, such as those described herein. The computer and/or IC may compute any of a variety of parameters or values that relate to backlight control and display mapping processes described herein. The image and video embodiments may be implemented in hardware, software, firmware and various combinations thereof.

Certain implementations of the invention comprise computer processors which execute software instructions which cause the processors to perform a method of the invention. For example, one or more processors in a display, an encoder, a set top box, a transcoder or the like may implement methods related to backlight control and display mapping processes as described above by executing software instructions in a program memory accessible to the processors. The invention may also be provided in the form of a program product. The program product may comprise any non-transitory medium which carries a set of computer-readable signals comprising instructions which, when executed by a data processor, cause the data processor to execute a method of the invention. Program products according to the invention may be in any of a wide variety of forms. The program product may comprise, for example, physical media such as magnetic data storage media including floppy diskettes, hard disk drives, optical data storage media including CD ROMs, DVDs, electronic data storage media including ROMs, flash RAM, or the like. The computer-readable signals on the program product may optionally be compressed or encrypted.

Where a component (e.g. a software module, processor, assembly, device, circuit, etc.) is referred to above, unless otherwise indicated, reference to that component (including a reference to a “means”) should be interpreted as including as equivalents of that component any component which performs the function of the described component (e.g., that is functionally equivalent), including components which are not structurally equivalent to the disclosed structure which performs the function in the illustrated example embodiments of the invention.

Equivalents, Extensions, Alternatives and Miscellaneous

Example embodiments that relate to efficient backlight control and display mapping processes are thus described. In the foregoing specification, embodiments of the present invention have been described with reference to numerous specific details that may vary from implementation to implementation. Thus, the sole and exclusive indicator of what is the invention, and is intended by the applicants to be the invention, is the set of claims that issue from this application, in the specific form in which such claims issue, including any subsequent correction. Any definitions expressly set forth herein for terms contained in such claims shall govern the meaning of such terms as used in the claims. Hence, no limitation, element, property, feature, advantage or attribute that is not expressly recited in a claim should limit the scope of such claim in any way. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense.

Claims

1. A method comprising: SceneCut = Δ_t ( Δ_t + c ),

accessing with a processor an input video stream comprising video frames;

accessing values of a function of at least one of minimum (min), average (mid), or maximum (max) luminance values of the video frames in the input video stream;

for a current frame in the input video stream: determining a probability measure of whether the current frame is in a new scene, wherein determining said probability measure comprises computing

where c is a tunable constant, SceneCut denotes the probability measure of a scene, and Δt denotes a weighted sum of respective Euclidian distances between luminance, chromaticity, and edge-strength histogram values of the current frame and luminance, chromaticity, and edge-strength histogram values of a prior frame; applying a temporal filter to the value of the function of the at least one of min, mid, or max luminance values of the current frame to generate a filtered output value, wherein one or more parameters of the temporal filter are based at least on the probability measure; determining an instantaneous dynamic range value for a target display based on the filtered output value and one or more limit luminance values for the target display; determining a backlight level value for the target display based on the instantaneous dynamic range value; and generating an output tone-mapped frame by mapping the instantaneous dynamic range value of the current frame to the target display operating at the backlight level value.

2. The method of claim 1, wherein the temporal filter comprises an exponential moving average filter.

3. The method of claim 2, wherein computing an output St with the exponential moving average filter comprises:

S0=LF(0), for t=0

St=α※LF(t)+β※St-1, for t>0

where α (alpha) and β (beta) denote weight factors and LF(t) denotes a function of the at least one of min, mid, or max luminance values for the current frame.

4. The method of claim 3, wherein the weight α is based on the probability measure, such that α is decreased if a frame has low probability to belong to a scene cut and is increased otherwise.

5. The method of claim 3, wherein

LF(t)=MedF(t),

wherein MedF(t) denotes the average (med) luminance value for the current frame.

6. The method of claim 3, wherein

β=1-α, and α=min(1,SceneCut※|St−1-LF(t)|※αscene+αbase),

wherein SceneCut is a scaler between 0 and 1 related to detecting a scene cut, and αscene and αbase are filter parameters.

7. The method of claim 3, wherein

α=αdef+(αmax−αdef) SceneCut,

where αdef denotes a default value, αmax denotes a maximum value, and SceneCut is a scaler between 0 and 1 related to detecting a scene cut.

8. The method of claim 3, wherein instead of initializing the EMA filter based on the luminance values of the current frame, the EMA filter is initialized based on the luminance values of a past scene, wherein the past scene is statistically similar to the current scene.

9. The method of claim 1, wherein determining an instantaneous dynamic range for the target display comprises computing one or more of where St denotes the filtered output value for the current frame, MinBL and MaxBL, denote the limit luminance values for the target display, and ƒmax(St), and ƒmin(St) are functions of the filtered output value and the limit luminance values for the target display.

MaxT=clip3(MinBL,ƒmax(ST), MaxBL),

MinT=clip3(MinBL,ƒmin(St), MaxBL),

10. The method of claim 9, wherein computing ƒmax(St) and ƒmin(St) comprises computing: wherein

log(St)=clip3(MinBL+w, log(St), MaxBL−w),

ƒmax(St)=log(St)+w,

ƒmin(St)=log(St)−w,

w denotes one half of the instantaneous dynamic range of the target display.

11. The method of claim 1, further comprising storing as metadata of the output tone-mapped frame the filtered output value of the current frame.

12. The method of claim 1, wherein the at least one of min, mid, or max luminance values are received as metadata of the input video stream.

13. The method of claim 1, wherein the at least one of min, mid, or max luminance values are generated by an image analysis processor based on luminance values of the video frames in the input video stream.

14. The method of claim 1, further comprising storing as metadata of the output tone-mapped frame the instantaneous dynamic range for the current frame for the target display.

15. An apparatus comprising a processor and configured to perform the method recited in claim 1.

16. A non-transitory computer-readable storage medium having stored thereon computer-executable instruction for executing a method with one or more processors in accordance with claim 1.